Optimizing Web Service Performance: A Comparative Analysis of Load Balancing Strategies Using NGINX and HAProxy with StoRM WebDAV Deployment.

In the rapidly evolving landscape of web services, achieving scalability, reliability, and performance has become imperative. Load balancing plays a pivotal role in distributing incoming network traffic across multiple servers to prevent overload, reduce response times, and maintain high availability. This thesis focuses on comparing the performance of two prominent load balancers, NGINX and HAProxy, in the context of StoRM WebDAV—a critical service for large-scale scientific data management developed by the Italian National Institute for Nuclear Physics (INFN). The study involves deploying StoRM WebDAV instances behind load balancers configured with NGINX and HAProxy to evaluate their effectiveness in optimizing performance. NGINX, a versatile web server and reverse proxy, is compared to HAProxy, a specialized load balancer known for its robustness and advanced traffic management features. Through detailed practical implementations, the research examines the integration, configuration, and performance of these tools in handling HTTP GET and PUT requests. Performance testing, including stress tests using tools like Vegeta, reveals insights into download speeds, resource utilization, and scalability. Results demonstrate trade-offs between direct access to StoRM WebDAV and access through load balancers. While NGINX introduces additional latency due to processing overhead, HAProxy achieves near-direct-access speeds with efficient CPU usage, making it preferable for high-speed data transfers. This thesis provides a comprehensive analysis of architecture design, system configuration, and performance metrics, offering valuable insights for optimizing StoRM WebDAV deployments. The findings contribute to the broader field of distributed systems by guiding developers and system administrators in selecting load-balancing strategies tailored to large-scale scientific and cloud-based environments.